As far as I can follow it, the BEA drift group (who as you would expect have done the most sophisticated drift analysis) have attempted a composite wind and current analysis. They mention the galley (which would have been wind affected) as well as a body at varying % submerged (with varying wind effect). I am unclear what they mean when the chart is titled "Example of reverse-drift simulation". Is this one of their actual solutions for AF447, or is it simply a theoretical example? They seem to suggest that the initial drift would have been to the S or SE - is this why they have been looking so far north?
Anyway, whatever this particular graph represents they have certainly attempted to incorporate both components (wind and current). Despite the sophistication of this analysis, there are a lot of unknowns, estimates and assumptions that would have gone into this. So we may have a very sophisticated analysis of incorrect inputs. And so the solution may be completely wrong - despite all the effort that went into this.
I believe that the only chance for drift analysis to give a meaningful answer is to be highly selective about what information is put into the model. Look for the purest signal and use the simplest model. Discard all the random noise - no one can know the balance of wind versus current for much of the debris. More data is not always better.
So we are back to the bodies - relying on the fact that we have 50 similar objects, and they will be almost entirely current affected and so we can eliminate the wind as a variable.
Takata and Fluid Flow have both adopted this approach for these very reasons. They have both come up with an impact point to the south. Takata assumed a constant current for 5 days prior to discovery (obviously a major assumption), which put the wreckage well to the south. Fluid Flow applied a regression analysis to the same data, and came up with a much closer point but in the same direction. This position is more likely than Takata's for three reasons:
a) it is hard to understand how the plane ended up as far south as Takata determined. It is even harder to imagine that it would have done so without their being some communication from the plane.
b) FluidFlow's answer incorporates the pollution spot (naturally, as he used this datum in this analysis)
c) Fluid Flow's solution also satisfies GreatBear's contention that the aircraft can't be very far from LKP.
One minor unknown in the body drift analysis. We know the date and location, but don't know the time of day that the body was discovered. Takata indicated that the first body found on June 6 was found at 0900Z. However he did not indicate the time that the bodies were discovered on subsequent days. He appeared to use the northernmost body and reverted to days (7,8,9 and 10 June). This detailed information may not be in the public domain. Knowing this information might smooth out some of this data and help refine the solution. We can probably assume they were found during daylight only - searching for bodies over an open ocean is a daylight task. We will have to assume a time if this information is not available. 1400Z (ie noon local time which would be close to the middle of the search period each day) is probably the best time to use. This will be closer to reality than assuming 2400Z. I don't know what Takata did for the locations on the subsequent days. He may have kept using 0900Z so he was working in 24 hr increments, or he may have used 0000 (the start of the day) or 2400 (the end of the day). In the absence of information, he had to make some assumption. Whatever assumption he made will have influenced his estimate of the average current speed.
The other refinement to Takata's analysis would be to use some form of "average" position for that day rather than the northernmost body. As will as being statistically more correct than relying on the extreme (most northern) datum each day, this would tie in with using 1400Z as the average time found for that day. This would reduce his estimated strength of the current, which would bring his impact point somewhere north of 2N (perhaps close to FluidFlow's solution).
If anyone has the raw data of the locations and the discovery times of the individual bodies, it would be interesting to play around with this.